developers.home-assistant/docs/development_testing.md

85 lines
4.1 KiB
Markdown

---
title: "Testing your code"
---
As it states in the [Style guidelines section](development_guidelines.md) all code is checked to verify the following:
- All the unit tests pass
- All code passes the checks from the linting tools
Local testing is done using [Tox](https://tox.readthedocs.io), which has been installed as part of running `script/setup` in the [virtual environment](development_environment.md). To start the tests, activate the virtual environment and simply run the command:
```shell
tox
```
It might be required that you install additional packages depending on your distribution/operating system:
- Fedora: `sudo dnf -y install systemd-devel gcc-c++`
- Ubuntu: `sudo apt-get install libudev-dev`
**Important:** Run `tox` before you create your pull request to avoid annoying fixes.
Running `tox` will run unit tests against the locally available Python releases, as well as validate the code and document style using `pycodestyle`, `pydocstyle` and `pylint`. You can run tests on only one `tox` target -- just use `-e` to select an environment. For example, `tox -e lint` runs the linters only, and `tox -e py38` runs unit tests only on Python 3.8.
`tox` uses virtual environments under the hood to create isolated testing environments. The `tox` virtual environments will get out-of-date when requirements change, causing test errors. Run `tox -r` to tell `tox` to recreate the virtual environments.
macOS users may see an `Error creating virtualenv` when runnng `tox`. If this occurs, install the [tox-venv](https://pypi.org/project/tox-venv/) package using the command `pip install tox-venv` and try again.
### Adding new dependencies to test environment
If you are working on tests for an integeration and you need the dependencies available inside the `tox` environment, update the list inside `script/gen_requirements_all.py`. Then run the script and then run `tox -r` to recreate the virtual environments.
### Running single tests using `tox`
You can pass arguments via `tox` to `py.test` to be able to run single test suites or test files. Replace `py38` with the Python version that you use.
```shell
# Stop after the first test fails
$ tox -e py38 -- tests/test_core.py -x
# Run test with specified name
$ tox -e py38 -- tests/test_core.py -k test_split_entity_id
# Fail a test after it runs for 2 seconds
$ tox -e py38 -- tests/test_core.py --timeout 2
# Show the 10 slowest tests
$ tox -e py38 -- tests/test_core.py --duration=10
```
### Testing outside of Tox
Running `tox` will invoke the full test suite. Even if you specify which tox target to run, you still run all tests inside that target. That's not very convenient to quickly iterate on your code! To be able to run the specific test suites without `tox`, you'll need to install the test dependencies into your Python environment:
```shell
pip3 install -r requirements_test_all.txt -c homeassistant/package_constraints.txt
```
Now that you have all test dependencies installed, you can run tests on individual files:
```shell
flake8 homeassistant/core.py
pylint homeassistant/core.py
pydocstyle homeassistant/core.py
py.test tests/test_core.py
```
You can also run linting tests against all changed files, as reported by `git diff upstream/dev... --diff-filter=d --name-only`, using the `lint` script:
```shell
script/lint
```
### Preventing linter errors
Save yourself the hassle of extra commits just to fix style errors by enabling the Flake8 git commit hook. Flake8 will check your code when you try to commit to the repository and block the commit if there are any style errors, which gives you a chance to fix them!
```shell
pip3 install flake8 flake8-docstrings
flake8 --install-hook=git
```
The `flake8-docstrings` extension will check docstrings according to [PEP257](https://www.python.org/dev/peps/pep-0257/) when running Flake8.
### Notes on PyLint and PEP8 validation
If you can't avoid a PyLint warning, add a comment to disable the PyLint check for that line with `# pylint: disable=YOUR-ERROR-NAME`. Example of an unavoidable one is if PyLint incorrectly reports that a certain object doesn't have a certain member.